Search Images Maps Play YouTube News Gmail Drive More »
Sign in
Screen reader users: click this link for accessible mode. Accessible mode has the same essential features but works better with your reader.

Patents

  1. Advanced Patent Search
Publication numberUS6614438 B1
Publication typeGrant
Application numberUS 09/583,028
Publication dateSep 2, 2003
Filing dateMay 30, 2000
Priority dateMay 30, 2000
Fee statusPaid
Also published asCN1331096C, CN1386245A, DE60104549D1, DE60104549T2, EP1290635A1, EP1290635B1, WO2001093198A1
Publication number09583028, 583028, US 6614438 B1, US 6614438B1, US-B1-6614438, US6614438 B1, US6614438B1
InventorsBernard Bru
Original AssigneeKoninlijke Philips Electronics N.V.
Export CitationBiBTeX, EndNote, RefMan
External Links: USPTO, USPTO Assignment, Espacenet
Data-processing arrangement for processing different types of data
US 6614438 B1
Abstract
A data-processing arrangement [DPA] is capable of processing different types of data [I, II, III] such as, for example, luminance (Y) and chrominance (UV) pixels from various pictures. The data-processing arrangement [DPA] comprises a controller [MCP] for executing a set of program items [PI]. A program item [PI] causes the data-processing arrangement [DPA] to fetch and process [F&PR] a block of data [DB] of a certain type [I, II, III]. The data-processing arrangement [DPA] further comprises an interface [INT] for temporarily storing a block of data [DB] in one of a plurality of memory sections [MS]. The memory section in which the block of data [DB] is stored is defined by the position that the program item [PI] for the block of data [DB] has within in the set of program items.
Images(8)
Previous page
Next page
Claims(5)
What is claimed is:
1. A data-processing arrangement (DPA) for processing different types of data (I, II, III), wherein the data processing arrangement (DPA) comprises:
a controller (MCP) for executing a set of program items (PI), a program item (PI) causing the data-processing arrangement (DPA) to fetch and process (F&PR) a block of data (DB) of a certain type (I, II, III); and
an interface (INT) for temporarily storing a block of data (DB) in one of a plurality of memory sections (MS), the memory section in which the block of data (DB) is stored being defined by the position that the program item (PI) for the block of data (DB) has within in the set of program items and not according to the type of data block (DB);
wherein the controller executes the set of program items (PI) in a predefined scheme and the position of the program item defines where the blocks of data to be processed are stored in the interface.
2. A method of configuring a data-processing arrangement (DPA) for processing different types of data (I, II, III), the method comprising:
a configuration step in which a controller (MCP) is programmed to execute a set of program items (PI), so that in response to a program item (PI), the controller (MCP) causes the data-processing arrangement (DPA) to fetch and process (F&PR) a block of data (DB) of a certain type (I, II, III) whereby the block of data is temporarily stored in one of a plurality of memory sections (MS), the memory section (MS) in which the block of data (DB) is stored being defined by the position that the program item (PI) for the block of data (DB) has within in the set of program items and not according to the type of data block (DB);
wherein in an execution step the controller (MCP) executes the set of program items (PI) in a predefined scheme and the position of the program item defines where the blocks of data (DB) to be processed are stored in the interface.
3. A computer program product for a data-processing arrangement (DPA) comprising a controller (MCP), the computer program product comprising a set of instructions which, when loaded into the controller (MCP), enables the controller (MCP) to execute a set of program items (PI) so that, in response to a program item (PI), the controller (MCP) causes the data-processing arrangement (DPA) to fetch and process (F&PR) a block of data (DB)) of a certain type (I, II, III) whereby the block of data (DB) is temporarily stored in one of a plurality of memory sections (MS), the memory section (MS) in which the block of data (DB) is stored being defined by the position that the program item (PI)for the block of data (DB) has within in the set of program items and not according to the type of data block (DB);
wherein in an execution step the controller (MCP) executes the set of program items (PI) in a predefined scheme and the position of the program item defines where the blocks of data (DB) to be processed are stored in the interface.
4. A computer program product for a data-processing arrangement (DPA), the computer program product comprising a set of program items (PI), a program item (PI), when executed, causing the data-processing arrangement (DPA) to fetch and process (F&PR) a block of data (DB) of a certain type (I, II, III) whereby the block of data is temporarily stored in one of a plurality of memory sections (MS), the memory section (MS) in which the block of data (DB)is stored being defined by the position that the program item (PI) for the block of data (DB) has within in the set of program items and not according to the type of data block (DB);
wherein in an execution instruction, the controller (MCP) executes the set of program items (PI) in a predefined scheme and the position of the program item defines where the blocks of data (DB) to be processed are stored in the interface.
5. A computer program product as claimed in claim 4, the computer program product comprising a program item [PI] which, when executed, causes the data-processing arrangement [DPA] to write a new portion of the set of program items [PI] into an internal control memory [CMEM].
Description
FIELD OF THE INVENTION

The invention relates to a data-processing arrangement for processing different types of data. The invention may be applied in, for example, a video-processing arrangement. A video-processing arrangement generally processes luminance (Y) and chrominance (UV) pixels, which are different types of data. The pixels that are to be processed may also differ in the sense that they originate from different pictures.

BACKGROUND OF THE INVENTION

A data-processing arrangement generally comprises an interface via which it receives data to be processed. Generally, the interface will temporarily store the data to be processed. Such storage may be necessary because, for example, the data-processing arrangement cannot always immediately process a data that it receives.

The interface may store a data which it receives in one or another memory section depending on the type of the data. For example, let it be assumed that the interface receives video comprising luminance and chrominance pixels. The interface may then have a memory section in which it stores luminance pixels and another memory section in which it stores chrominance pixels. The video to be processed may originate from different pictures because, for example, the processing includes a picture-in-picture (PIP) feature. In that case, there may be different groups of memory sections, each group of memory sections belonging to a different picture.

SUMMARY OF THE INVENTION

It is an object of the invention to allow cost-efficiency.

According to the invention, a data-processing arrangement, which is capable of processing different types of data, comprises a controller for executing a set of program items. Each program item causes the data-processing arrangement to fetch and process a block of data of a certain type. An interface temporarily stores a block of data in one of a plurality of memory sections. The memory section in which the block of data is stored is defined by the position that the program item for the block of data has within in the set of program items.

The invention takes the following aspects into consideration. A data-processing arrangement may have to process various composites of data. One composite may comprise a relatively high percentage of certain type of data whereas another composite may comprise a relatively low percentage of that data. For example, video data may comprise relatively many chrominance pixels, or relatively few, depending on the video format.

A data-processing arrangement will generally receive a composite of data in the form of a stream that is a time multiplex of different types of data. The stream may comprise a relatively high percentage of a certain type of data during one time interval whereas it comprises a relatively low percentage of that data during an other time interval.

According to the background art, the interface temporarily stores data in one or another memory section depending on the type of data. That is, each type of data has its own memory section. A memory section should be sufficiently large so that it can temporarily store the type of data which it has been assigned to store. In view of what has been explained hereinbefore, the memory section may need to have a relatively large size for one composite of data whereas it can have a relatively small size for another composite. Moreover, it may occur that a memory section needs to store relatively much data during a certain time interval, while another memory section has to store relatively few data, whereas the inverse applies in another time interval.

There are basically two solutions to cope with these aspects that are related to the background art. One solution is that each memory section has a relatively large size such that it can always store the type of data which it has been assigned to store. This solution is relatively expensive because it requires much storage capacity which, on average, will not be used efficiently. The other solution is to dynamically divide a memory into various sections, each memory section belonging to a certain type of data. A memory section for a certain type of data is made larger or smaller depending on the storage capacity that is required for that type of data. This solution is also relatively expensive because it requires a memory-management system that is relatively complicated in terms of hardware or software, or both.

According to the invention, a controller executes a set of program items, a program item causing the data-processing arrangement to fetch and process a block of data of a certain type. The interface temporarily stores the block of data in one or another memory section depending on the position that the program item has within the set of program items.

The controller will execute the program items in accordance with a pre-defined scheme. Since, according to the invention, a program-item causes a block of data to be temporarily stored in one or another memory section depending on the position that the program item has within the set of program items, the latter defines a scheme in accordance with which blocks of data are temporarily stored in memory sections of an interface. Consequently, there is no need for a complicated memory-management system for managing storage of data in the interface. This contributes to cost-efficiency.

The invention also allows cost-efficiency for the following reason. The data-processing arrangement will successively process blocks of data. This implies that the interface will successively apply blocks of data to one or more processing circuits. According to the invention, a program items causes a block of data to be processed and the position of the program item defines where the block of data is stored in the interface. Consequently, at any given instant, it is immediately clear where a particular block of data, which needs to be processed at that instant, is stored in the interface. Thus, the hardware or software, or both, which causes blocks of data to be read from the interface so that they are processed, can be relatively simple.

Furthermore, the invention allows an efficient use of the memory sections that are present in the interface. According to the invention, a program item causes the data-processing arrangement to fetch and process a certain block of data. This implies that there is a data-fetch delay. The data-fetch delay is the difference between the instant when a data block is written into a memory section and the instant when the data block is read because it will enter a processing pipeline. Preferably, the interface is provided with just enough memory sections to bridge the data-fetch delay, not more and not less. In that case, the memory sections will be filled with data most of the time and, consequently, they will be used efficiently.

The memory sections can have a uniform size. In many applications, it will be possible to suitably divide up the data to be processed into blocks so that most blocks will have a size that corresponds with the size of the memory sections. This further contributes to an efficient use of the memory sections.

These and other aspects of the invention will be described in greater detail hereinafter with reference to drawings.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 is a conceptual diagram illustrating basic features of the invention described hereinbefore;

FIG. 2 is a block diagram illustrating a video-processing arrangement in accordance with the invention;

FIG. 3 is a block diagram illustrating a first filter block of the video-processing arrangement;

FIG. 4 is a block diagram illustrating an output block of the video-processing arrangement;

FIG. 5 is a diagram illustrating an example of how the video-processing arrangement processes a block of pixels;

FIG. 6 is a block diagram illustrating a control register of the video-processing arrangement;

FIG. 7 is a diagram illustrating the basic operational scheme of the video-processing arrangement;

FIG. 8 is a block diagram illustrating a controller of the video-processing arrangement;

FIG. 9 is a diagram illustrating a portion of a scheduler for the controller;

FIG. 10 is a diagram illustrating a data-fetch file for the controller;

FIG. 11 is a diagram illustrating a series of steps that controller carries out;

FIG. 12 is a block diagram illustrating a main-bus interface of the video-processing arrangement.

EMBODIMENT OF THE INVENTION

The following remarks relate to reference signs. Like entities are designated by like letter references in all Figures. Several similar entities may appear in a single Figure. In that case, a digit or a suffix is added to the letter reference in order to distinguish like entities. The digit or the suffix may be omitted for convenience or it may be replaced by an asterisk in the case where its value is not important (do not care value). This applies to the description as well as the claims.

FIG. 1 illustrates basic features of the invention as outlined hereinbefore. A data-processing arrangement [DPA] is capable of processing different types of data [I, II, III]. The data-processing arrangement [DPA] comprises a controller [MCP] for executing a set of program items [PI]. A program item [PI] causes the data-processing arrangement [DPA] to fetch and process [F&PR] a block of data [DB] of a certain type [I, II, III]. The data-processing arrangement [DPA] further comprises an interface [INT] for temporarily storing a block of data [DB] in one of a plurality of memory sections [MS]. The memory section in which the block of data [DB] is stored is defined by the position that the program item [PI] for the block of data [DB] has within in the set of program items.

The features illustrated in FIG. 1 may be applied in, for example, a video-processing arrangement. The video-processing arrangement may be used, for example, to modify video data from one to another display format, for example by modifying the number of lines or the number of pixels per line, or both. The video-processing arrangement may also be used, for example, to insert a sub-picture or graphics, or both, into a main picture.

FIG. 2 illustrates a video-processing arrangement in accordance with the invention. The video-processing arrangement receives video data from a main memory, which is not shown, via a main bus [HWY]. In response, it provides a video-output stream [VIDOUT] that may be applied to, for example, a video display device or a video-recording device. The video-processing arrangement comprises a main-bus interface [MBIF], two filter blocks: a first and second filter block [XPB1, XPB2], an output block [XPC], a streaming-out circuit [SO] and a controller [MCP]. The two filter blocks [XPB] and the output block [XPC] are coupled to the main-bus interface [MBIF] via two internal busses: a first and second internal bus [XBUS1, XBUS2]. The controller [MCP] is coupled to the various other blocks shown in FIG.2 via a control bus [CBUS]. There are data paths that allow data transfers between the two filter blocks [XPB] and the output block [XPC]. These data paths are illustrated in FIG. 2 but not named.

FIG. 3 illustrates the first filter block [XPB1]. The second filter block [XPB2] is identical. The first filter block [XPB1] comprises two internal-bus interfaces [IBIF], one for each internal bus [XBUS] and a data-transfer circuit [TRNSFR]. The data-transfer circuit [TRNSFR] allows a transfer of data to the other filter block, which is the second filter block [XPB2], and the output block [XPC]. The first filter block [XPB1] further comprises an input cross bar [XBARI], an output cross bar [XBARO], 24 memory circuits in the form of a first to twenty-fourth SRAM [SRAM1-SRAM24], and four processing circuits in the form of a first and second horizontal filter [HFIL1, HFIL2] and a first and second vertical filter [VFIL1, VFIL2]. The input cross bar [XBARI] may receive data from the other filter block, which is the second filter block [XPB2], or from the output block [XPC].

FIG. 4 illustrates the output block [XPC]. The output block [XPC] comprises two internal-bus interfaces [IBIF], one for each internal bus [XBUS] and a data-transfer circuit [TRNSFR]. The output block [XPC] further comprises two input cross bars: a first and second input cross bar [XBARI1, XBARI2], an output cross bar [XBARO], 24 memory circuits in the form of a first to twenty-fourth SRAM [SRAM1-SRAM24], and five processing circuits in the form of a matrix/dematrix circuit [MD], two upsamplers [UPS], a blender [BLND] and a look-up table [LUT]. The first and the second input cross bar [XBARI1, XBAR2] may receive data from the two filter blocks [XPB]. The second input cross bar [XBARI2] may also receive data from the controller [MCP]. The output cross bar [XBARO] may apply data to the streaming-out circuit [SO] and the controller [MCP].

The video-processing arrangement illustrated in FIG. 2 basically operates as follows. The main-bus interface [MBIF] fetches a block of pixels from the main memory, which is not shown in FIG. 3, via the main bus [HWY]. Subsequently, the main-bus interface [MBIF] transfers the block of pixels to an SRAM in one of the filter blocks [XPB] or in the output block [XPC] via one of the internal busses [XBUS]. Subsequently, one of the processing circuits in the filter blocks [XPB], or the output block [XPC], processes the block of pixels. Accordingly, a once-processed block is obtained. It is stored in another SRAM. Subsequently the same or another processing circuit may process the once-processed block of pixels. This step-by-step processing continues until a completely processed block of pixels is obtained that is stored in an SRAM in the output block [XPC]. Subsequently, the output block [XPC] transfers the completely processed block of pixels to the streaming-out circuit [SO].

The video-processing arrangement carries out the processing described hereinbefore successively for different blocks of pixels. Consequently, the streaming-out circuit [SO] receives successive completely processed blocks of pixels. It casts the pixels it receives in an appropriate format and adds some synchronization signals. Accordingly, the video-output stream [VIDOUT] is obtained.

FIG.5 illustrates an example of the processing of a block of pixels. It comprises three processing steps: an (i)-th, (i+1)-th and (i+2)-th processing step [PS(i), PS(i+1), PS(i+2)]. It assumed that the main-bus interface [MBIF] shown in FIG.2 has already fetched the block of pixels from the main memory. The block of pixels is therefore present in the main-bus interface [MBIF].

In the (i)-th processing step [PS(i)], the main-bus interface [MBIF] transfers the block of pixels to the first SRAM [SRAM1] in the first filter block [XPB1] via the first internal bus [XBUS1]. In the (i+1)-th processing step [PS(i+1)], the first filter block [XPB1] processes the block of pixels. More specifically, the first horizontal filter [HFIL1] reads the block of pixels from the first SRAM [SRAM1] pixel-by-pixel, it processes the pixels that it reads and it writes processed pixels in the twelfth SRAM [SRAM12] of the output block [XPC]. This transfer goes via the output crossbar [XBARO] and the data-transfer circuit [TRNSFR] of the first filter block [XPB1] illustrated in FIG.3, and via the first input crossbar [XBARI1] of the output block [XPC] illustrated in FIG. 4. Accordingly, the twelfth SRAM [SRAM12] of the output block [XPC] will contain a processed block of pixels at the end of the (i+1)-th processing step [PS(i+1)]. In the (i+2)-th processing step [PS(i+2)], the output block [XPC] transfers the processed block of pixels to the streaming-out circuit [SO].

The processing illustrated in FIG. 5 corresponds to a video-processing chain that is composed of a single horizontal filter. The video-processing arrangement illustrated in FIG. 2 allows many different video-processing chains. For example, it allows a video-processing chain that comprises a sequence of filters, some of which may be in parallel, and other types of processing such as, for example, blending, up- or down sampling. The SRAMs, which are comprised in the two filter blocks [XPB] and the output block [XPC], serve as FIFO memories between two successive elements in a video-processing chain, an element being a certain processing operation.

In principle, the video-processing arrangement illustrated in FIG. 2 can begin to process a new block of pixels each processing step. Thus, the video-processing arrangement can process different blocks of data simultaneously during a certain processing step. For example, referring to FIG. 6, in the (i+1)-th processing step [PS(i+1)], the main-bus interface [MBIF] may transfer a new block of pixels to an SRAM of the first or second filter block [XPB1, XPB2] or the output block [XPC]. The new block of pixels may undergo, for example, the video-processing chain illustrated in FIG. 5. In that case, the operations illustrated in FIG. 5 shift by one processing step for the new block of pixels.

Moreover, the video-processing arrangement illustrated in FIG. 2 can begin to process two new blocks of pixels each processing step. This is related to the presence of the two internal busses [XBUS]. Each internal bus [XBUS] allows a transfer of a block of pixels from the main-bus interface [MBIF] to the first or second filter block [XPB1, XPB2] or the output block [XPC] within a single processing step. For example, referring to FIG. 5 and the (i)-th processing step [PS(i)] illustrated therein, while the main-bus interface [MBIF] transfers the block of pixels concerned to the first SRAM [SRAM1] of the first filter block [XPB1] via the first internal bus [XBUS1], it may transfer another block of pixels to another SRAM in the first filter block [XPB1] or an SRAM in the second filter block [XPB2] or the output block [XPC] via the second internal bus [XBUS2].

The controller [MCP] shown in FIG. 2 causes the video-processing arrangement to process a block of pixels in accordance with a certain video-processing chain. This processing involves a series of processing steps [PS] as described hereinbefore with reference to FIG. 5. Furthermore, the processing in a certain processing step [PS] involves only a certain subset of circuits present in the video-processing arrangement. For each processing step, the controller [MCP] suitably programs the subset of circuits that is involved in that processing step.

For example, let us consider the processing illustrated in FIG. 5 and, more specifically, the processing in the (i+1)-th processing step [PS(i+1)]. It involves the following circuits of the first filter block [XPB1] illustrated in FIG.3: the first SRAM [SRAM1], the input cross bar [XBARI], the first horizontal filter [HFIL1], the output cross bar [XBARO] and the transfer block [TRNSFR]. The processing also involves the following circuits of output block [XPC] illustrated in FIG. 4: the first input cross bar [XBARI1] and the twelfth SRAM [SRAM12].

The controller [MCP] shown in FIG. 3 programs the aforementioned circuits so that, in combination, they carry out the processing illustrated in the (i+1)-th processing step [PS(i+1)] shown in FIG. 5 and described hereinbefore. The controller [MCP] programs the first SRAM [SRAM1] of the first filter block [XPB1] so that it provides the block of pixels that need to be filtered. It may do so, for example, by defining an appropriate read start-address. The controller programs the input cross bar [XBARI] of the first filter block [XPB1] so that it couples the first [SRAM1] with the first horizontal filter [HFIL1]. The controller programs the first horizontal filter [HFIL1] so that it carries out the desired type of filtering. The controller [MCP] programs the output cross bar [XBARO] and the data-transfer circuit [TRNSFR] of the first filter block [XPB1] and the first input cross bar [XBARI1] of the output block [XPC] so that these circuits couple the first horizontal filter [HFIL1] of the first filter block [XPB1] with the twelfth SRAM [SRAM12] of the output block [XPC]. Finally, the controller [MCP] programs the twelfth SRAM [SRAM12] in output block [XPC] so that the filtered pixels are appropriately stored. It may do so, for example, by defining an appropriate write start-address.

The circuits of the two filter blocks [XPB] illustrated in FIG. 3 and the output block [XPC] illustrated in FIG. 4 have control registers. The content of a control register defines the functional behavior of the circuit to which the control register belongs. Consequently, the controller [MCP] shown in FIG. 3 programs the circuits in the filter blocks [XPB] and the output block [XPC] by writing control data into the respective control registers. The controller [MCP] can also program a circuit by managing the control data that are already contained in the control register of that circuit.

FIG.6 illustrates a control register [CREG]. The control register [CREG] comprises a front register [FREG], a shadow register [SREG] and a register controller [RCNTRL]. It basically operates as follows. The control register [CREG] receives control data from the controller [MCP] shown in FIG. 2 via the control bus [CBUS] also shown in FIG.2. This control data is first written into the front register [FREG]. In response to a control command from the controller [MCP], the register controller [RCNRTL] may cause the control data to be transferred to the shadow register [SREG]. The control data that is contained in the shadow register [SREG] defines the functional behavior of the circuit to which the control register belongs. The control data that is contained in the shadow register [SREG] may be transferred back to the front register [FREG] when the control data that is contained in the front register [FREG] is transferred to the shadow register [SREG]. The register controller [RCNTRL] causes this swap operation to be carried out in response to a control command from the controller [MCP].

FIG. 7 illustrates the basic operational scheme of the video-processing arrangement illustrated in FIG. 2. The horizontal dimension represents time in clock cycle units. The video-processing arrangement alternately carries out processing steps [PS] and configuration steps [CS].

In a processing step [PS], the video-processing arrangement manipulates one or more blocks of pixels. For example, a block of pixels may be passed through a processing circuit or simply transferred from one memory circuit to another memory circuit. In parallel, another block of pixels may be passed through another processing circuit or transferred to yet another memory circuit. These manipulations of blocks of pixels are defined by the control data that are contained in the shadow registers of the respective circuits in the video-processing arrangement.

In a processing step [PS], the controller [MCP] shown in FIG. 2 may write control data into various control registers while data is being processed. It further instructs the respective register controllers what action they should carry out in the configuration step [CS] that will follow. The respective control data in the shadow registers is not modified because these control data define the data manipulations that are being carried out. The writing of control data into the front registers during a processing step can be seen as a preparation for the subsequent processing step. That is, while data is being manipulated during an (n)-th processing step [PS(n)], the data manipulations to be carried in the subsequent (n+1)-th processing step [PS(n+1)] are defined.

In a configuration step [CS], the content of the respective shadow registers can be modified. For example, referring to the control register [CREG] illustrated in FIG. 7, the register controller [RCNTRL] can cause the shadow register [SREG] to take over the control data that is contained in the front register [FREG]. The register controller [RCNTRL] may also leave the control data in the shadow register [SREG] unaltered. Thus, a configuration step [CS] merely involves a transfer of control data within the respective control registers, if there is any such transfer. Moreover, control data can be transferred simultaneously in the respective control registers. Consequently, a configuration step will require a few clock cycles only.

The processing steps [PS] and the configuration steps [CS] alternately succeed one another in the following manner. Let it be assumed that the video-processing circuit illustrated in FIG. 2 carries out a processing step. This generally implies that various processing circuits process blocks of pixels. A processing circuit that has processed its block of pixels signals the controller [MCP] shown in FIG. 2 that it has completed its job. The controller [MCP] initiates a configuration step when it has received such a ready signal from each processing circuit that was active. Likewise, a control register that has transferred control data as instructed signals to the controller [MCP] that it has completed its job. This implies that the appropriate control data has been placed in the shadow register. The controller [MCP] initiates a processing step [PS] when it has received such a configuration-ready signal from each control register in which a control-data transfer took place.

FIG.8 illustrates the controller [MCP] of the video-processing arrangement. The controller [MCP] comprises a control processor [CPU] and a control memory [CMEM]. The control memory [CMEM] contains the following items: a kernel [KRNL], a library of subroutines [SR], a scheduler [SCHED], a data-fetch file [DFF] and a task-continuation file [TCF]. The control memory [CMEM] may also contain further items of data such as, for example, variables. The kernel [KRNL] and the subroutines [SR] comprise instructions for the control processor [CPU]. The scheduler [SCHED], the data-fetch file [DFF] and the task-continuation file [TCF] comprise addresses of subroutines.

A sub-routine [SR] defines a certain manipulation of a block of data during a processing step [PS]. For example, FIG. 5 illustrates that in the (i+1)-th processing step [PS(i+1)] a block of pixels is read from the first SRAM [SRAM1] in the first filter block [XPB1] so as to be processed by the first horizontal filter [HFIL1] in the first filter block [XPB1], the processed block of pixels being written into the twelfth SRAM [SRAM12] in the output block [XPC]. This manipulation can thus be defined by a means of a sub-routine [SR]. In terms of video signal processing, a sub-routine [SR] relates to a certain element of a video-signal processing chain.

More specifically, a sub-routine [SR] causes the control processor [CPU] to program the video-processing arrangement such that the requested manipulation will be carried out. That is, a sub-routine [SR] causes the control processor [CPU] to write control data into the control register [CREG] of the circuits that are involved in the manipulation, or to move control words already contained in these control register [CREG], or both, so that the shadow registers [SREG] of these circuits contain appropriate control data.

FIG. 9 illustrates a portion of the scheduler [SCHED]. The scheduler [SCHED] is shown as matrix having rows and columns. FIG. 9 shows a part of the scheduler [SCHED] only: a (j)-th and (j+1)-th row [R(j), R(j+1)] and a (k)-th to (k+7)-th column [C(k), C (k+7)]. The part of the scheduler [SCHED] that is shown concerns one line period of a picture. Each cell defines a certain task for the video-processing arrangement. A task concerns a block of data that is contained in the main memory. The number of the task, 0, 1, 2, 3 or 4, defines what type of data is concerned and what the video-processing arrangement should do with this data in terms of data processing and transfers.

For example, task number 1 may concern a block of luminance pixels for a high-definition picture. Task number 2 may concern a block of chrominance pixels for the same high-definition picture. Task number 3 may concern a block of luminance pixels for a standard-definition picture. Each different block of pixels should be processed in accordance with a different video signal-processing chain. The task number 4 may concern new data for the scheduler [SCHED]. It consists of updating the scheduler [SCHED] with this new data. Task number 0 (zero) is a so-called empty task. An empty task does not relate to a block of data and, consequently, it does not cause the video-processing arrangement to transfer or process any data.

More specifically, a cell that defines a non-empty task (1,2,3,4) contains a start address of a sub-routine. This sub-routine is the first of a chain subroutines that causes the video-processing arrangement to carry out the required task. A sub-routine whose start address is contained in the scheduler [SCHED] will therefore be called first sub-routine hereinafter. Each sub-routine of a chain is carried out in a different processing step.

The chains of subroutines are formed and managed by means of the following items contained in the control memory [CMEM] shown in FIG. 8: the kernel [KRNL], the data-fetch file [DFF] and the task-continuation file [TCF]. That is, these items cause the control processor [CPU] to control the video-processing arrangement so that it properly carries out the various tasks defined in the scheduler [SCHED]. This will now be explained in greater detail.

Referring to FIG. 2, a first sub-routine instructs the main-bus interface [MBIF] to fetch the block of data concerned from the main memory. In addition, the first sub-routine will write a start address of a second sub-routine into the data-fetch file [DFF] shown in FIG. 8. The second sub-routine will cause the main-bus interface [MBIF] to transfer the block of data to a memory circuit in the first or second filter block [XPB1, XPB2] or the output block [XPC]. This transfer will take place via the first or second internal bus [XBUS1, XBUS2] depending on whether the start-address of the first sub-routine was in the (j)-th or (j+1)-th row [R(j), R(j+1)], respectively, of the scheduler [SCHED] illustrated in FIG. 9. The second sub-routine should be executed only when it is certain that the block of data concerned is present in the main-bus interface [MBIF]. To that end, the second sub-routine is executed after a certain delay. This delay is implemented by means of the data-fetch file [DFF] in the control memory [CMEM] shown in FIG. 8.

FIG. 10 illustrates the data-fetch file [DFF]. The data-fetch file [DFF] can functionally be regarded as a collection of three registers in the form of tables: an initial, intermediate and final highway-fetch register [HFP0, HFP1, and HFP2]. Each highway-fetch register [HFP] corresponds to 4 consecutive processing steps. More specifically, each row of a highway-fetch register [HFP] corresponds to a processing step.

The data-fetch file [DFF] functionally behaves as follows. In an (n)-th processing step, the control processor [CPU] shown in FIG. 8 executes the first subroutines that are defined in the cells on the (j)-th and (j+1)-th row [R(j), R(j+1)] and in the (k)-th column [C(k)] of the scheduler [SCHED] illustrated in FIG. 10. These first subroutines will cause the control processor [CPU] to write the start addresses of the respective second subroutines in the first row of the initial highway-fetch register [HFP0]. In a subsequent (n+1)-th processing step, the control processor [CPU] executes the first subroutines that are defined in the cells on the (j)-th and (j+1)-th row [R(j), R(j+1)] and in the (k+1)-th column [C(k+1)] of the scheduler [SCHED] illustrated in FIG. 10. These first subroutines will cause the control processor [CPU] to write start addresses of the respective second subroutines in the second row of the initial highway-fetch register [HFP0].

The procedure described in the preceding paragraph continues processing step by processing step. Consequently, the initial highway-fetch register [HFP0] will be completely filled in an (n+4)-th processing step. Its content is then copied to the intermediate highway-fetch register [HFP1] after the content of the intermediate highway-fetch register [HFP1] has been copied to the final highway-fetch register [HFP2]. In effect, the highway-fetch registers [HFP] are shifted one position to the left each cycles of 4 processing steps. In an (i+5)-th processing step, the control processor [CPU] shown in FIG.8 executes the first subroutines that are defined in the cells on the (j)-th and (j+1)-th row [R(j), R(j+1)] and in the (k+5)-th column [C(k+5)] of the scheduler [SCHED] illustrated in FIG. 10. Since the initial highway-fetch register [HFP0] has been emptied, these first subroutines will cause the control processor [CPU] to write the start addresses of the respective second subroutines in the first row of the initial highway-fetch register [HFP0].

The control processor [CPU] shown in FIG. 8 reads start addresses of respective second subroutines from the final highway-fetch register [HFP2] so as to execute these subroutines. More specifically, when the control processor [CPU] has written a start address of a second sub-routine in a cell of the initial highway-fetch register [HFP0], it subsequently reads a start address of a second sub-routine from the cell in the final highway-fetch register [HFP2] that has the same position. The control processor [CPU] will then execute the latter second sub-routine whose start address it has read from the final highway-fetch register [HFP2]. It has already been explained the highway-fetch registers [HFP] are shifted one position to the left each cycle of 4 processing steps.

Consequently, the data-fetch file [DFF] illustrated in FIG. 10 causes a delay of 8 processing steps between the first sub-routine of a task, which is a request for data from the main memory, and the second sub-routine of the task, which is the transfer of this data from the main-bus interface [MBIF] shown in FIG. 3 to the first or second filter block [XPB1, XPB2] or the output block [XPC] also shown in FIG. 2. This delay accounts for a latency, which accompanies data fetches via the main bus [HWY] and the main-bus interface [MBIF]. This latency, as well as the main-bus interface [MBIF], will be described in greater detail hereinafter.

When the control processor [CPU] shown in FIG. 8 has carried out the first and second sub-routine of a task, the block of data which is the subject of that task will be present in a memory circuit of the first or second filter block or the output block. The video-processing arrangement is then ready to process the block of data, or just transfers it, by means of a third and further subroutines. These subroutines are executed as follows.

The second sub-routine causes the control processor [CPU] shown in FIG. 8 to write the start address of the third sub-routine in the task-continuation file [TCF] also shown in FIG. 8. This writing is done within the processing step wherein the second sub-routine is executed. In the subsequent processing step, the control processor [CPU] reads the start address of the third sub-routine and carries out this sub-routine. This causes the control processor [CPU] to write the start address of a fourth sub-routine, if any, in the task-continuation file [TCF]. In the then subsequent processing step, the control processor [CPU] reads the start address of the fourth sub-routine and carries out this sub-routine. This causes the control processor [CPU] to write the start address of a fifth sub-routine, if any, in the task-continuation file [TCF]. This procedure continues until the last sub-routine of the task has been carried out. The task has then been completed.

Referring to FIG. 8, the kernel [KRNL] is the program that causes the control processor [CPU] to read start addresses of subroutines in the scheduler [SCHED], the data-fetch file [DFF] and the task-continuation file [TCF] and to carry out these subroutines as described hereinbefore. The kernel [KRNL] can be activated by means of an interrupt that is generated in the streaming-out circuit [SO] shown in FIG. 2. This interrupt indicates that the video-output stream [VIDOUT] has reached a certain point in a current frame. A frame typically comprises some blanking lines before the lines to be displayed. The streaming-out circuit [SO] may apply an interrupt to the control processor [CPU] so as to activate the kernel [KRNL] when, for example, the video-output stream [VIDOUT] has reached the last-but-one blanking line. The kernel [KRNL] may be stopped after the task that concerns the last block of pixels of a frame. The scheduler [SCHED] illustrated in FIG. 9 preferably comprises a special task which appropriately stops all processing and resets the circuits in the video-processing arrangement.

FIG. 11 illustrates a series of steps that the kernel [KRNL] causes the control processor to carry out each processing step. It is assumed, of course, that the kernel [KRNL] is active. The series of steps comprise a first to fifth step [S1-S5].

In the first step [S1], the control processor reads two cells in the scheduler [SCHED] illustrated in FIG. 9 [RD(SCHED)]. The two cells that are read depend on the current processing step and the current line. Referring to FIG. 9, the control processor moves one column to the right every processing step that belongs to a certain line. The control processor moves two rows downwards every new line. A line can be defined in terms of a certain number of processing steps. For example, in FIG. 9, a line comprises 8 processing steps. It has already been explained that a cell of the scheduler [SCHED] defines a tasks in the form of a start addresses of a first sub-routine of a chain of subroutines. Thus, in the first step [S1], the control processor reads two first-subroutine start addresses and writes these start addresses into a subroutine-execution file. This file may be, for example, in the form of an assembly of registers.

In the second step [S2], the control processor reads two cells on a row in the final highway-fetch register [RD(HFP2)]. The two cells that are read depend on the current processing step. Referring to FIG. 10, the control processor moves one row downwards every processing step until the last row has been reached. The next row will then be the top row of the final highway-fetch register [HFP2]. It has already been explained that the cells of the final highway-fetch register [HPF2] contain start addresses of second subroutines in a task. Thus, in the second step [S2], the control processor reads two second-subroutine start addresses and writes these start-addresses into the subroutine-execution file.

In the third step [S3], the control processor reads a pointer in the streaming-out circuit [SO] shown in FIG. 2 [RD(SO)]. This pointer indicates the fullness of a buffer in which processed pixels are temporarily stored before the are outputted in the form of video-output stream [VIDOUT]. The pointer tells the control processor [CPU], as it were,.whether the streaming-out circuit [SO] needs further processed pixels or not.

In the fourth step [S4], the control processor reads the start addresses that are contained in the task-continuation file [TCF] shown in FIG. 8. These start addresses concern third and further subroutines in a task. That is, the task-continuation file [TCF] contains the start addresses of the subroutines other than the first and second sub-routine of a task. The control processor writes the start addresses that it reads from the task-continuation file [TCF] into the addresses into the subroutine-execution file.

In the fifth step [S5], the control processor reads the start addresses that are contained in the subroutine-execution file one after the other. In response, the control processor executes the subroutines that go with these start addresses one after the other. Thus, it can be said that, for each processing step, the control processor makes a list of subroutines to be executed in the first to fourth step [S1-S4] and that it actually executes the sub-routines which are in this list in fifth step [S5].

The following aspect of the kernel [KRNL] is not shown in FIG. 10. In principle, the video-processing arrangement illustrated in FIG. 2 processes successive blocks of pixels faster than the streaming-out circuit [SO] outputs pixels in the form of the video-output stream [VIDOUT]. Thus, the video-processing arrangement should interrupt its processing from time to time so as to prevent an overflow in the streaming-out circuit [SO]. The video-processing arrangement can be made to interrupt its processing by temporarily deactivating the kernel [KRNL].

The kernel [KRNL] comprises the following features that prevent the video-processing arrangement illustrated in FIG. 2 from being too much ahead, as it were, with respect to the video-output stream [VIDOUT]. Each cycle of 4 processing steps, the kernel [KRNL] causes the control processor [CPU] shown in FIG. 8 to read a register in the streaming-out circuit [SO] that indicates the position of the current pixel that will be outputted. This register thus indicates whether the processing in the video-processing arrangement is too much ahead, as it were, with respect to the video-output stream [VIDOUT] or not. If the video-processing arrangement is too much ahead, the kernel [KRNL] is de-activated for a certain period of time

FIG. 12 illustrates the main-bus interface [MBIF] of the video-processing arrangement shown in FIG. 2. It comprises two buffer memories: a first and second buffer memory [IRAM1, IRAM2] each having a size of 2 Kbytes. Each buffer memory [IRAM] is divided into 8 blocks: a first to eight block [B1, . . . B8] each having a size of 256 bytes.

The main-bus interface [MBIF] basically operates as follows. The buffer memories [IRAM] are alternately coupled to the main bus [HWY] and to the internal busses [XBUS] each cycle of 4 processing steps. That is, the first buffer memory [IRAM1] will be coupled to the main bus [HWY] during a period of 4 consecutive processing steps whereas the second buffer memory [IRAM2] is coupled to the internal busses [XBUS]. This situation is illustrated in FIG. 12. The inverse will apply for the subsequent period of 4 processing steps: the second buffer memory [IRAM2] will be coupled to the main bus [HWY] whereas the first buffer memory [IRAM1] will be coupled to the internal busses [XBUS]. The buffer memory [IRAM1/IRAM2] which, for a given period of 4 consecutive processing steps, is coupled to the main bus [HWY] serves to receive data from the external main memory. Meanwhile, the other buffer memory [IRAM2/IRAM1], which is coupled to the internal busses [XBUS], serves to transfer data that has been received from the main memory to the first or second filter block or the output block. Thus, with each cycle of 4 consecutive processing steps the buffer memories [IRAM] change roles.

A transfer of data from the main memory to the first or second filter block or the output block takes 3 cycles of 4 consecutive processing steps. In the first cycle, the controller requests the main-bus interface [MBIF] to fetch the data from the main memory. In the second cycle, the data is transferred from the main memory to the buffer memory [IRAM1/IRAM2] that is coupled to the main bus [HWY]. In the third cycle, the same buffer memory [IRAM1/IRAM2] is coupled to the internal busses [XBUS] and the data is transferred from the buffer memory [IRAM1/IRAM2] to the first or second filter block or the output block. It is noted that there is a correspondence with the data-fetch file. The first, second and third cycle described hereinbefore correspond to the initial data-fetch register [HPFO], the intermediate data-fetch register [HFP1] and the final data-fetch register [HFP2], respectively, which registers are illustrated in FIG. 10.

The main-bus interface [MBIF] stores data in the following manner. A buffer memory [IRAM] corresponds to a cycle of 4 consecutive processing steps. The main-bus interface [MBIF] stores in the first and second block [B1, B2] the data that will be transferred from the buffer memory [IRAM] to the first or second filter block or the output block in the first processing step of this cycle. More specifically, it stores the data that will be transferred via the first internal bus [XBUS1] in the first block [B1] and the data that will be transferred via the second internal bus [XBUS2] in the second block [B2]. The data that will be transferred via the first and second internal bus [XBUS1, XBUS2] in the second processing step of the cycle concerned are stored in the third and fourth block [B3, B4], respectively, and so on.

In summary, the first and second block [B1, B2], the third and fourth block [B3, B4], the fifth and sixth block [B5, B6], and the seventh and eight block [B7, B8] correspond to a first, second, third and fourth processing step, respectively, in a cycle of 4 processing steps. The blocks that have an odd number contain data that will be transferred via the first internal bus [XBUS1]. The blocks that have an even number contain data that will be transferred via the second internal bus [XBUS2]. Consequently, each cell of the scheduler [SCHED] illustrated in FIG. 9 is associated with a specific block [B] in a buffer memory [IRAM]. Since a cell of the scheduler [SCHED] defines a particular block of data that needs to be manipulated by the video-processing circuit, it is predefined where this block of data will be stored in the main-bus interface [MBIF].

The scheduler [SCHED] shown in FIG. 8 can be written into the control memory [CMEM] of the controller [MCP] in the following manner. The scheduler [SCHED] illustrated in FIG. 9 may comprise one or more cells that define a task for updating the scheduler [SCHED]. It has already been mentioned that the task number 4 illustrated in FIG. 9 may concern new data for the scheduler [SCHED]. Task number 4 will cause the video-processing arrangement to read from the main memory a block of data that defines a new part of the scheduler [SCHED] and to transfer this block of data from the main memory to the control memory [CMEM].

The main memory may contain the scheduler in a compressed form. In principle, any data compression technique can be used to compress the data that is contained in the scheduler. For example, the portion of the scheduler [SCHED] illustrated in FIG. 9 comprises a series of subsequent cells that contains task number 1. Consequently, the scheduler [SCHED] can be compressed by indicating the series of cells that contain task number 1 instead of defining task number 1 for each cell separately. The control processor [CPU] shown in FIG. 8 can be programmed to decompress a block of data that defines a new portion of the scheduler in a compressed form. Thus, task number 4 illustrated in FIG. 9 may not only cause a block of data that defines a new portion of the scheduler to be read from the main memory. It may also cause the block of data to be decompressed and subsequently added to the scheduler.

The video-processing arrangement illustrated in FIG. 2 can be physically implemented and, subsequently, made operational in the following manner. In a first step, the video-processing arrangement is manufactured, for example, in the form of an integrated circuit containing the various blocks shown in FIG. 3.

In a second step, various control-software items are loaded into the control memory [CMEM], which is shown in FIG. 8. The kernel [KRNL] and the subroutines [SR] are examples of such control-software items. A further control-software item may be, for example, an initialization program that causes a first portion of a scheduler [SCHED] to be loaded into the control memory [CMEM]. The initialization program may further create the outlines of the data-fetch file [DFF] and the task-continuation file [TCF], for example, by defining suitable memory ranges for these items. The control-software items, which are loaded into the control memory [CMEM], may be contained on a data carrier such as, for example, a disk or a non-volatile memory. The control-software items may be retrieved from such a data carrier via a communication network such as, for example, the Internet.

In a third step, one or more schedulers [SCHED] are loaded into a memory that is coupled to the video-processing arrangement illustrated in FIG. 3 via the main bus [HWY]. A scheduler [SCHED] is a set of program-software items. The control-software items, which have been loaded into the control memory [CMEM], enable the controller [MCP] to execute the program-software items contained in the scheduler [SCHED]. The controller [MCP] will read the scheduler [SCHED] and execute the program items contained therein. A program-software item, when executed, causes the video-processing arrangement to fetch and process a certain block of data. Consequently, the video-processing arrangement will successively process blocks of data. These blocks of data may be contained in the same memory as the scheduler [SCHED] or in separate memory. The scheduler [SCHED], which is loaded into the memory that is coupled to the main bus [HWY], may be contained on a data carrier such as, for example, a disk or a non-volatile memory. The scheduler [SCHED] may be retrieved from such a data carrier via a communication network such as, for example, the Internet.

The video-processing arrangement illustrated in FIG. 2 and further described with reference to FIG. 3 to FIG. 12 is an example of an implementation of the basic features illustrated in FIG. 1. The program items [PI] illustrated in FIG. 1 take the form of the data contained in the scheduler [SCHED] illustrated in FIG. 9. It has been explained with reference to FIG. 9 to FIG. 12 that the controller [MCP] executes the scheduler [SCHED]. It has further been explained that an execution of a program item causes the video-processing arrangement illustrated in FIG. 2 to fetch and process a block of data. The interface illustrated in FIG. 1 takes the form of the main-bus interface [MBIF] illustrated in FIG. 12. The main-bus interface [MBIF] comprises two buffer memories [IRAM1, IRAM2]. Each buffer memory is divided into eight blocks [B1-B8]. Consequently, a memory section [MS] illustrated in FIG. 1 takes the form of a block [B] in FIG. 12. The main-bus interface [MBIF] temporarily stores a block of data in one of the blocks [B]. The block [B] in which the block of data is stored, is defined by the position of the data in the scheduler [SCHED] that has caused the block of data to be fetched so as to be processed.

The drawings and their description hereinbefore illustrate rather than limit the invention. It will be evident that there are numerous alternatives, which fall within the scope of the appended claims. In this respect, the following closing remarks are made.

There are numerous ways of implementing functions by means of items of hardware or software, or both. In this respect, the drawings are very diagrammatic, each representing only one possible embodiment of the invention. Thus, although a drawing shows different functions as different blocks, this by no means excludes that a single item of hardware or software carries out several functions. Nor does it exclude that an assembly of items of hardware or software or both carry out a function.

Any reference sign in a claim should not be construed as limiting the claim. The word comprising does not exclude the presence of other elements or steps than those listed in a claim. The word a or an preceding an element or step does not exclude the presence of a plurality of such elements or steps.

Patent Citations
Cited PatentFiling datePublication dateApplicantTitle
US6081895 *Oct 10, 1997Jun 27, 2000Motorola, Inc.Method and system for managing data unit processing
US6173389 *Dec 4, 1998Jan 9, 2001Billions Of Operations Per Second, Inc.Methods and apparatus for dynamic very long instruction word sub-instruction selection for execution time parallelism in an indirect very long instruction word processor
US6314568 *Jan 6, 1999Nov 6, 2001Nec CorporationBroadcast-program viewing method and system to allow customized viewing based on user input
USRE34850 *Dec 6, 1991Feb 7, 1995Mitsubishi Denki Kabushiki KaishaDigital signal processor
Referenced by
Citing PatentFiling datePublication dateApplicantTitle
US7050324 *Jul 13, 2004May 23, 2006Fulcrum Microsystems, Inc.Asynchronous static random access memory
US7161828Aug 9, 2005Jan 9, 2007Fulcrum Microsystems, Inc.Asynchronous static random access memory
US7612781Sep 20, 2005Nov 3, 2009Sony Computer Entertainment Inc.Memory control method of graphic processor unit
US7657679 *Oct 13, 2006Feb 2, 2010Via Technologies, Inc.Packet processing systems and methods
US7760205Jul 25, 2005Jul 20, 2010Sony Computer Entertainment Inc.Information processing apparatus for efficient image processing
US8370557Dec 19, 2008Feb 5, 2013Intel CorporationPseudo dual-port SRAM and a shared memory switch using multiple memory banks and a sideband memory
EP1647887A2 *Aug 8, 2005Apr 19, 2006Sony Computer Entertainment Inc.Apparatus for efficient image processing
EP1647888A2 *Aug 10, 2005Apr 19, 2006Sony Computer Entertainment Inc.Apparatus for efficient image processing
WO2005008672A2 *Jul 13, 2004Jan 27, 2005Uri CummingsAsynchronous static random access memory
WO2006033226A1 *Aug 30, 2005Mar 30, 2006Sony Comp Entertainment IncMemory control method of graphic processor
Classifications
U.S. Classification345/520, 345/522, 715/762, 712/E09.047
International ClassificationH04N5/14, G06T1/60, G06F12/02, H04N5/262, G06F12/00, G06F9/38
Cooperative ClassificationG06T1/60
European ClassificationG06F9/38D2, G06T1/60
Legal Events
DateCodeEventDescription
Dec 20, 2011ASAssignment
Effective date: 20110824
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:IPG ELECTRONICS 503 LIMITED;REEL/FRAME:027497/0001
Owner name: FUNAI ELECTRIC CO., LTD., JAPAN
Feb 24, 2011FPAYFee payment
Year of fee payment: 8
Feb 20, 2007FPAYFee payment
Year of fee payment: 4
May 29, 2003ASAssignment
Owner name: KONINKLIJKE PHILIPS ELECTRONICS N.V., NETHERLANDS
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:U.S. PHILIPS CORPORATION;REEL/FRAME:013687/0860
Effective date: 20030522
Owner name: KONINKLIJKE PHILIPS ELECTRONICS N.V. GROENOWOUDSEW
Aug 24, 2000ASAssignment
Owner name: U.S. PHILIPS CORPORATION, NEW YORK
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:BRU, BERNARD;REEL/FRAME:011145/0772
Effective date: 20000627